What the work actually is

You pick a task from a pre-validated pool: a C repository pinned to a specific commit, arriving with a draft engineering question, positive and negative rubrics, foils, and a golden solution. Your job is to go deeper than the draft. You read the subsystem, trace the relevant PR and issue history, and often build and run the thing — with gdb, a sanitizer, strace or a profiler — until you can state what happens at runtime rather than what the source appears to say. Then you rewrite the question so that a frontier coding agent given repository access still gets it wrong, while a competent senior engineer with the same access would arrive at the golden answer and agree it was fair.

The rubric work is where most of the craft lives. Foils have to be plausible-but-wrong in the specific way a model goes wrong — the confident misreading of a macro-expanded branch, the assumption that a `#ifdef` path is the one that compiles on the target platform, the answer that describes the API contract instead of the implementation. You then run the full Studio validation loop (Check, Golden, Run, Analyze) and iterate until every stage passes, document in the task README, and respond to reviewer feedback on anything returned.

What the screen is looking for

Substantial production C, not C-adjacent C++ or a university course. Specifically: evidence you have landed in an unfamiliar large C codebase and figured out its runtime behaviour under preprocessor-heavy builds, platform conditionals, manual memory management, and autotools/CMake/Meson build graphs. Depth in databases, cryptography, network protocols or systems programming is weighted heavily — the task mix leans toward architecture and system design (39%) and code onboarding (34%), with a meaningful security slice. Expect follow-ups that test whether you can name the tool, the flag, and what the output told you. Written precision matters more than usual here, because the questions and rubrics you write are the deliverable.

Logistics

  • Fully remote and asynchronous; expert team leads sit across US, India and Nigeria time zones.
  • Work happens in Mercor Studio; time is tracked with Insightful and access is provisioned through Okta.
  • Onboarding gates: identity verification, background check, signed Terms of Work and CIIAA, a tax form, and a short calibration quiz before task access opens.
  • Pay is per approved task at an observed $130; approval requires a second engineer and a super reviewer, so throughput early on is lower than it looks while you learn the rubric bar.